I think this is a bit of a shortsighted take. The fundamental question is: should Google allow its systems/software to be used to generate homophobic/sexist/racist/transphobic content? If Google can determine with high likelihood (perhaps through a combination of language modeling of your Google Docs/email + examining your browsing history, search history, email contacts, purchase history, physical location, etc) that you are generating that sort of content, why should you be allowed to use Google for anything?
Home Depot generally has no way to know how you're using their products once they're sold. The very nature of an internet service means that Google continuously monitors your usage of their products.
But that isn't at all what Google is trying to do here. The first example in the Twitter thread is an "inclusiveness warning" about the word landlord.
On the one hand, nobody is forced to use their suggested words just like nobody is forced to fix the grammar mistakes Google underlines. So it may have some cultural influence but it isn't really authoritarian. It's also not creepy in the sense that it is not using your data outside of what is directly in the Doc.
On the other hand, it is not flagging actually bigoted content, it is more akin to GitHub renaming the master branch (or at the very least that is a subset of what this flags). It would be horrifying if that kind of "infraction" actually did prevent you from using Google services. And I'd be seriously skeptical of any attempted bigotry detection like you're describing that is built on top of the system discussed here.
